City-Scale Multi-Camera Vehicle Tracking of Vehicles based on YOLOv7
Duong Nguyen‐Ngoc Tran, Long Hoang Pham, Huy-Hung Nguyen, Jae Wook Jeon
Abstract
Multi-Target Multi-Camera Tracking has a massive Intelligent Traffic Surveillance System application domain. In this paper, we introduce the framework using YOLOv7, DeepSORT, and Re-ID to provide the identification of vehicles via a camera network. At first, we localize a vehicle in every single camera. Then, the Re-ID obtains the unique feature of each bounding box we crop from detection. We use DeepSORT and Trajectory Clustering from the list of features to keep the vehicle's identity from a single camera to a multi-camera. The experiment result shows the promise of the framework in the extensive system.
Topics & Concepts
Computer scienceComputer visionArtificial intelligenceMinimum bounding boxVehicle tracking systemTracking (education)Cluster analysisTrajectoryIdentification (biology)Camera auto-calibrationFeature (linguistics)Domain (mathematical analysis)Stereo cameraCamera resectioningImage (mathematics)MathematicsAstronomyPsychologyMathematical analysisBiologyBotanySegmentationLinguisticsPhysicsPedagogyPhilosophyVideo Surveillance and Tracking MethodsAdvanced Neural Network ApplicationsAdvanced Image and Video Retrieval Techniques